The purpose of the present paper is to study how the statistical dispersion of distortion and biomechanical parameters may be used in producing an objective evaluation of voice quality. For such, the behaviour of the same GMM classifiers used in the detection of pathology will be exploited. The work will show specific cases derived from a database of normal and pathological voice, set into contrast against a Universal Background Model built from the population of normal subjects. Results will be contrasted against classical subjective scoring and a proposal for automatic voice quality evaluation in terms of the most relevant parameters will also be discussed.

Index Terms. GRBAS, Voice Pathology Grading, Gaussian Mixture Models